Paperonium

← Back to the plain-language case

Reported renal-cancer P=0.004 is P=0.24 under the paper's own stated Fisher test

JAMA (2010) · doi:10.1001/jama.2010.1877 · PMC3655326

Summary

The paper compares the prevalence of cancers in Cowden or Cowden-like (CS/CSL) patients with germline KILLIN promoter methylation against patients with germline PTEN mutations. The Abstract and the Results both report renal cell carcinoma in 4/45 methylation-positive patients and 6/155 PTEN-mutation-positive patients, with P=0.004. The Methods say these prevalences were compared with a two-tailed Fisher exact test. On those counts that test gives P=0.237 (odds ratio 2.42; exact 95% CI 0.48–10.72), and none of the other standard tests for a 2×2 table gives a value below P=0.09. The renal association is one of the two cancer risks named in the abstract Conclusions, and the Comment bases advice on kidney surveillance and test prioritization on it. The paper's primary finding (methylation in 45/123 patients) and its breast-cancer comparison are not affected.

What the paper reports

Abstract, Results:

Individuals with KILLIN-promoter methylation had 3-fold increased prevalence of breast cancer (35/42 vs 24/64, P<0.0001) and a >2-fold increase of kidney cancer (4/45 vs 6/155, P=0.004) over those with germline PTEN mutations.

Abstract, Conclusions:

Germline KILLIN- methylation is common among CS/CSL patients, and associated with increased risks of breast and renal cancer over PTEN mutation-positive patients. These observations need to be replicated.

Results, paragraph "Prevalent Cancers in Pathogenic PTEN Mutation Positive versus KILLIN Methylation":

Renal cell carcinoma was over-represented in the methylation positive subjects over PTEN mutation positive individuals (4/45 vs 6/155, P=0.004).

Methods, "Statistical Analysis":

The prevalence of component malignancies between KILLIN promoter methylation positive patients and germline pathogenic PTEN mutation positive patients was compared using the Fisher’s 2-tailed exact test with P<0.05 considered to be significant.

Comment, second paragraph:

Patients with CS-like features, especially where breast cancer and/or renal carcinomas are present in the individual and/or family, should be offered KILLIN methylation analysis first

Comment, third paragraph:

In contrast to germline PTEN mutations, germline methylation of the KILLIN promoter confers a significantly higher prevalence of female invasive breast cancer and renal cell carcinomas. The current national practice guidelines for those with PTEN germline mutations includes heightened surveillance of the female breasts and thyroid, but do not have awareness of renal cancer risk. If our observations of two- to three-fold increased risks of renal and/or breast cancer with KILLIN germline methylation over those of PTEN mutation holds, then extra vigilance for the organs-at-risk, breast and kidneys, is warranted.

Comment, "Conclusions":

Among CS/CSL patients, presence of germline KILLIN gene promoter hypermethylation was common and was associated with increased risk of breast and renal cancer compared with PTEN mutation-positive patients.

Every quote was checked verbatim against the article as archived in PubMed Central (the author manuscript). The publisher's version of record gives the same counts and P value in JAMA house style ("4/45 vs 6/155; P = .004").

What the paper's own numbers show

Method. The reported counts give the 2×2 table [[4, 41], [6, 149]]: renal cell carcinoma present or absent among 45 methylation-positive and 155 PTEN-mutation-positive patients. I applied the test the Methods name (two-tailed Fisher exact), then the other common tests for this table, to see whether any of them gives 0.004. I also worked out which counts would be needed to reach P=0.004. As a control, I recomputed the breast-cancer comparison from the same sentence and the methylation prevalence the same way. Every value below is printed by recompute.py.txt (scipy 1.18.1). The paper has no tables holding these counts; they appear only in the text.

Quantity Location Reported Recomputed
Renal prevalence, methylated vs PTEN Abstract Results; Results 4/45 vs 6/155, ">2-fold" 8.9% vs 3.9%; prevalence ratio 2.30 (consistent)
Expected renal cases among the 45 if prevalence were equal derived — 2.25 (observed 4; 10 cases among 200 patients)
Fisher exact, two-tailed (the test the Methods name) Abstract Results; Results P=0.004 P=0.237
Fisher exact, one-tailed same counts — P=0.164
Fisher one-tailed mid-p same counts — P=0.106
Pearson chi-square, uncorrected / Yates-corrected same counts — P=0.174 / 0.331
Barnard / Boschloo exact, two-sided same counts — P=0.212 / 0.208
Exact binomial, 4/45 against a fixed rate of 6/155 (one-sided) same counts — P=0.096
Odds ratio (sample), Wald 95% CI, Wald P same counts not reported 2.42 (0.65–8.99), P=0.186
Odds ratio (conditional MLE), exact 95% CI same counts not reported 2.41 (0.48–10.72)
Renal cases among the 45 needed for two-tailed Fisher P≤0.004 (PTEN group held at 6/155) sensitivity 4 8 (P=0.0038)
Renal cases among the 45 needed for P<0.05 (PTEN group held at 6/155) sensitivity 4 6 (P=0.029)
PTEN-group renal cases needed for P≤0.004 (methylated group held at 4/45) sensitivity 6/155 0/155 (P=0.0023); 1/155 gives P=0.0096
Breast cancer, 35/42 vs 24/64 (control check) Abstract Results; Results P<0.0001 Fisher two-tailed P=3.1×10⁻⁶ (consistent)
Methylation prevalence (control check) Abstract Results 45/123, 37% (95% CI 29–45%) 36.6%; Wald 28.1–45.1%, exact 28.1–45.7% (consistent)

The reported P value is about 1/59 of the value the stated test gives. The ">2-fold" prevalence ratio is correct. Only the P value, and the claim of statistical significance that rests on it, do not follow from the counts.

A smaller point, which does not change any conclusion: the Abstract calls the breast-cancer difference a "3-fold increased prevalence", but 35/42 vs 24/64 is 83.3% vs 37.5%, a prevalence ratio of 2.22 (odds ratio 8.33).

Figure

Reported P value for renal cancer compared with P values recomputed from the paper's counts under eight tests

Figure 1. P values for renal cell carcinoma in 4/45 KILLIN-methylation-positive vs 6/155 PTEN-mutation-positive patients, on one logarithmic axis. The orange point is the value reported in the Abstract and Results (P=0.004). Blue points are recomputed from the same counts; the larger blue point is the two-tailed Fisher exact test named in the Methods (P=0.237), joined to the reported value by a dotted line. The dashed line marks the paper's significance threshold, P=0.05. Every recomputed value lies to the right of the threshold, between 0.096 and 0.331, so the conclusion does not depend on the choice of test.

Two-tailed Fisher P as a function of the number of renal cancers among the 45 methylation-positive patients

Figure 2. Two-tailed Fisher exact P (blue, log scale) for x/45 methylation-positive patients with renal cancer against the reported 6/155 in the PTEN-mutation group, for x from 0 to 12. At the reported count of 4/45 the test gives P=0.237. The reported P=0.004 (orange dashed line) is first reached at 8/45 (P=0.0038), twice the reported count. A count of 6/45 would be needed to cross P=0.05.

Why it matters

Explanations considered

None of these explanations accounts for the reported value.

Status

As of 2026-10-23:

Suggested correction

In the Abstract Results and in the Results paragraph on prevalent cancers, replace "P=0.004" for renal cell carcinoma (4/45 vs 6/155) with the two-tailed Fisher exact value, P=0.24, and preferably give the odds ratio with its exact 95% CI (2.41, 0.48–10.72). Then revise the statements that rest on this result: the abstract Conclusions, the Comment ("confers a significantly higher prevalence of ... renal cell carcinomas", and the kidney-surveillance and testing-priority statements) and the closing Conclusions. They could say that renal cancer was more common in the methylation-positive group (8.9% vs 3.9%) but that the difference was not statistically significant in this sample. If the counts, rather than the P value, are in error, the correction should give the correct counts and describe the 155-patient PTEN-mutation comparison group. Separately, the breast-cancer difference described as "3-fold" corresponds to a prevalence ratio of 2.2 (83.3% vs 37.5%).